I downloaded the release tag of fpc 3.0.0 via svn and built it using the 2.6.4 seed compiler. It seems to work OK.

But now I got to thinking:
Should I make fpc 3.0.0 once more using the newly created fpc 3.0.0? Will there be some difference between the two 3.0.0 versions if I do?

It shouldn't, that's why the compiler is built more than once in the standard makefile sequence.

Should I make fpc 3.0.0 once more using the newly created fpc 3.0.0?

Yes if you do compile by hand (without Makefile)

Will there be some difference between the two 3.0.0 versions if I do?

If what you mean is: does 3.0.0 compiler binary produced from single compilation by 2.6.4 differ from the one by 3.0.0? Then yes.

Else if what you mean is: will using 2.6.4 as the starting compiler produce different 3.0.0 binary from 3.0.0 while using the Makefile? Then no.

leledumbo wrote

> Should I make fpc 3.0.0 once more using the newly created fpc 3.0.0? Yes if you do compile by hand (without Makefile)

Which I never do...

...

Else if what you mean is: will using 2.6.4 as the starting compiler produce different 3.0.0 binary from 3.0.0 while using the Makefile? Then no.

I always use the make all command via the Makefile. So do I read you right that the Makefile compiles the compiler twice, once using the seed compiler and then using the thus created compiler?

Bo Berglund wrote

I always use the make all command via the Makefile. So do I read you right that the Makefile compiles the compiler twice, once using the seed compiler and then using the thus created compiler?

It compiles it thrice: once more in the end to ensure that the resulting compiler is "stable", in the sense that it generates the same code when compiled with itself as when compiled with a previous release.

Jonas

Fantastic! Then I can be assured a successful make results in the expected release version!
Good job!

But what about optimising the RTL to -O4?

I did a separate pass to build the RTL with -O4 which 2.64 does not support. I found it makes a useful difference to the execution speed of some programs.

Peter

There is a variant of the OPT parameter that applies options only for the last compilation. This way you can use options that the compiling compiler does not yet support. Now if I'd only remember that... -.- (for cross compilation one can use CROSS_OPT)

Regards,
Sven

In our previous episode, Sven Barth said:

There is a variant of the OPT parameter that applies options only for the last compilation. This way you can use options that the compiling compiler does not yet support. Now if I'd only remember that... -.- (for cross compilation one can use CROSS_OPT)

NEWOPT

"Marco van de Voort" wrote

In our previous episode, Sven Barth said: > There is a variant of the OPT parameter that applies options only for

the

> last compilation. This way you can use options that the compiling

compiler

> does not yet support. Now if I'd only remember that... -.- (for cross > compilation one can use CROSS_OPT) NEWOPT

Ah, thanks. :) (and CROSS_OPT might be CROSSOPT... :/ )

Regards,
Sven

Thanks Marco, thats useful to know. Can't find it mentioned anywhere in the Programers Guide.